Performance Tips and Tricks Mixed Precision Training Faster Image Processing libjpeg-turbo Pillow-SIMD Background Installation How to check whether you’re running Pillow or Pi...

Normalization Normalization When training a model, it helps if your input data is normalized—that is, has a mean of 0 and a standard deviation of 1. But most images and compute...

Conclusion Conclusion In this chapter we explored the last application covered out of the box by the fastai library: text. We saw two types of models: language models that can ...

From Data to DataLoaders Data Augmentation From Data to DataLoaders DataLoaders is a thin class that just stores whatever DataLoader objects you pass to it, and makes them a...

vision.models Computer Vision models zoo class Darknet [source] [test] class WideResNet [source] [test] wrn_22 [source] [test] vision.models Overview of the models ...

Regression Assemble the Data Training a Model Regression It’s easy to think of deep learning models as being classified into domains, like computer vision, NLP, and so forth....

Support Overview Reporting Issues Do’s and Don’ts PRs Support Overview fastai support is provided via github issue tracker and the forums . Most issues, in particular p...

Callbacks Creating a Callback Callback Ordering and Exceptions Callbacks Sometimes you need to change how things work a little bit. In fact, we have already seen examples of ...

Questionnaire Further Research Questionnaire Why do we first resize to a large size on the CPU, and then to a smaller size on the GPU? If you are not familiar with regular ex...

CAM and Hooks CAM and Hooks The class activation map (CAM) was introduced by Bolei Zhou et al. in “Learning Deep Features for Discriminative Localization” . It uses the output ...